GLM-5.3-Flash Pricing, Benchmarks & Channels
GLM-5.3-Flash is a 2026-08-26 model from Mistral AI featuring a 1M context window and extended reasoning capabilities. Lowest rate on OpenRouter ($0.0399 in / $0.0784 out per 1M tokens).
Input Price / 1M
$0.15
Output Price / 1M
$0.5
Cache Read / 1M
$0.03
Context Window
1M
Generation Speed
132 TPS
Access Grade
A · Global
F.A.R.E. Four-Dimension Score Breakdown
Scenario Capability Breakdown (0–100):
💻 Coding
88/100
✍️ Writing
84/100
📚 Long-Doc RAG
94/100
👁️ Vision OCR
87/100
F.A.R.E. CAPABILITY RADAR
GLM-5.3-Flash Multi-Platform Price Comparison (9 Verified Channels)
All prices are in USD per 1M tokens. Domestic developers can prioritize channels marked "Direct in CN" for compliant access and local billing.
| Platform / Gateway | Network Access | Input ($/1M) | Output ($/1M) | Cache Hit ($/1M) | Billing & Notes | Link |
|---|---|---|---|---|---|---|
| OpenRouter Lowest Price | Direct in CN | $0.0399 | $0.0784 | $0.03 | Direct CN access · RMB billing | Visit Official → |
| 302.AI Gateway | Direct in CN | $0.075 | $0.25 | $0.015 | Direct CN access · RMB billing | Visit Official → |
| Volcengine Ark (CN) | Direct in CN | $0.1187 | $0.4156 | $0.0341 | Direct CN access · RMB billing | Visit Official → |
| Zhipu BigModel (CN) | Direct in CN | $0.1195 | $0.4183 | $0.0344 | Direct CN access · RMB billing | Visit Official → |
| SiliconFlow (CN) | Direct in CN | $0.1199 | $0.4198 | $0.0345 | Direct CN access · RMB billing | Visit Official → |
| Mistral Official | Global Network | $0.15 | $0.5 | $0.03 | Global cloud billing | Visit Official → |
| Together AI | Global Network | $0.15 | $0.5 | $0.03 | Global cloud billing | Visit Official → |
| Fireworks AI | Global Network | $0.15 | $0.5 | $0.03 | Global cloud billing | Visit Official → |
| DeepInfra | Global Network | $0.15 | $0.5 | $0.03 | Global cloud billing | Visit Official → |
✅ Core Strengths & Ideal Workloads
- Top-tier empirical capability score (88 index)
- Fast generation speed (~132 tokens/sec)
- Ultra-low token cost, ideal for high-volume batch workloads
- Open weights available for private self-hosting (~20GB VRAM at INT4)
⚠️ Limitations & Practical Caveats
- European frontier lab available via Mistral La Plateforme, AWS Bedrock, and OpenRouter.
- Extended thinking generates internal reasoning tokens (~1.8x output multiplier)